The present invention relates generally to the testing of semiconductor chips, and specifically to the design of probe contactors for such testing prior to packaging.
Typically, semiconductor chips are tested to verify that they function appropriately and reliably. This is often done when the semiconductor devices are still in wafer form, that is, before they are diced from the wafer and packaged. This allows the simultaneous testing of many chips at a single time, creating considerable advantages in cost and process time compared to testing individual chips once they are packaged. If chips are found to be defective, then when the chips are diced from the wafer, the defective ones can be discarded and only the reliable chips are packaged. It is an axiom then that the larger a wafer that may be reliably tested at a time, the more savings can be incurred in cost and process time.
Generally, when performing wafer testing, a chuck carrying a wafer is raised to a probe card to which thousands of probes are electrically coupled. To test larger wafers, small, high performance probes are needed. The probes must be able to break through the oxide and debris layers on the surface of the contact pads of the chips on the wafer in order to make a reliable electrical contact to each pad. Additionally, the probes must be able to compensate for the fact that the contact pads may be of different heights (i.e., not all the contact pads on a wafer may reside in the same plane). Furthermore, the chuck and the probe card mechanical mount may not be precisely parallel and flat, introducing further height variations which the probes must accommodate.
Conventionally, cantilever wire probes have been used to test wafers in this regard. However, cantilever wire probes are too long and difficult to accurately assemble to allow reliable simultaneous contact to all of the chips on a conventional wafer. Additionally, cantilever wire probes have high self and mutual inductance problems which do not make them good candidates for testing of high-speed devices. These problems are exaggerated when they are used to test larger wafers. Cantilever (or bending) probes can also be fabricated at a small physical scale by various mircofabrication techniques known in the art. These cantilever springs lack the mechanical energy density (for controlled scrubbing of the oxide layer) and spatial efficiency to be ideally effective for reliable testing of large wafers.
A number of attempts have been made to overcome the deficiencies of cantilever probes, all with varied levels of success. For instance, U.S. Pat. No. 5,926,951, assigned to Form Factor, Inc., describes methods of producing spring probes by coating a ductile metal with a spring metal (as seen in FIG. 4). These springs are bending mode springs similar to cantilever springs. Moreover, they are elongated and poorly supported in lateral directions at the contactor causing problems with controlled scrubbing of the contact pads. Furthermore, probes that require a long spring length such as these have relatively poor electrical performance.
U.S. Pat. No. 6,426,638, assigned to Decision Track, describes a torsion spring design, see FIG. 1, which is more mechanically efficient than other spring designs, and more effective than cantilever designs. U.S. Pat. No. 6,771,084, also assigned to Decision Track, describes the fundamental principle of a single footed torsion spring probe contactor, see FIGS. 2A and 2B. However these designs too, have their limitations. The particular incarnations considered and contemplated by these patents do not address or solve many of the practical requirements for a spring probe contactor such as: range of motion, optical characteristics required for vision recognition, practical production means, requirements for high lateral stability of the contact tip in response to scrubbing forces to name a few.
Improvements in the design of probe contactors have come with advances in photolithography and associated micromachining techniques. U.S. Pat. No. 5,190,637 to Wisconsin Alumni Research Foundation describes the basis of multi-layer build up fabrication through lithographic electro-forming techniques of three-dimensional metal structures including springs and spring contactors. The present applicants have created a micro-formed torsion bar probe contactor which overcomes many of the deficiencies of the prior art and is a subject of the instant application.
Another aspect of the present application is the formation of the tip at the end of the probe. Older pin based contactors, such as cantilever needle probes or vertically buckling beam probes, are typically built from wire with a sharpened or shaped tip. This type of geometry provides for adequate electrical contact only if substantial contact force is applied. High contact force is deleterious to the semiconductor devices under test which often include active devices under the I/O pads. Furthermore, pin based contactors cannot be built at the fine pitches and high pin counts required for modern large wafer test. For these and other reasons, microfabricated probe contactors are an attractive alternative to pin based probe cards.
Microfabricated probe contact tips for use on contactor probes have been proposed in a variety of configurations and are plentiful in the art. In most of these configurations, provision is made for the creation of a tip with a well defined and controlled surface shape, size, material, and texture. Each of these elements is important for achieving the required consistent electrical contact to common IC pad metals such as Al, AlSiCu, Cu, Cu alloys, Au, or solder. Each of these parameters has a bearing on the contact performance but control over the geometry is among the most significant and is a function of the fabrication technology employed.
Another factor that is often overlooked is the optical characteristic of the tip and adjacent structures. Typically, probe cards are used in conjunction with wafer probers equipped with machine vision systems for automatic identification of probe tip locations and alignment of those to the I/O pads on the wafer, such as that described by U.S. Pat. No. 5,321,352, assigned to Tokyo Electron Labs. Basically, a machine vision system includes a camera that is positionable and looks at the tips of the probe needles. The camera has some magnification appropriate for viewing the geometry of the tip. It also includes a light source such as an LED ring light or a co-axial light. The image from the camera is processed by computer so as to determine the location of the tip relative to the camera's image area. This location information is used by the prober's computer control algorithm to position the device-under-test (DUT) bond pads accurately under the probe tips. Thus the probe tip must be designed with the vision systems requirements in mind. In particular, vision systems require a good optical contrast between the tip and adjacent structures, particularly in the case of microfabricated contactors with small physical dimensions between adjacent surfaces. Typical microfabricated spring contactors have smooth planar surfaces in close proximity to the contact tip surface, creating difficulty with regard to the vision recognition systems due to reflections from surfaces other than the tip, as seen in FIG. 5A. Thus, these vision systems often mistake unrelated structures for the tip causing vision rejections or errors in the captured tip position.
Various attempts to overcome this problem have been suggested, but each have had their own problems. For instance, U.S. Pat. No. 6,255,126, assigned to Form Factor, Inc. and shown in FIGS. 3A and 3B, discloses a pyramid shaped contactor tip for use with a cantilever probe structure. The pyramid is formed by replicating an anisotropically etched cavity in silicon and bonding the replicated tip to a spring structure. While this technique may produce a tip with a good mechanical strength due to its wide base, and the sides of the pyramid reflect light off-axis and appear dark under the normal illumination used for machine vision recognition, this design has at least two significant drawbacks. The fabrication sequence is driven by a mold replication technique and requires a separate bonding step in order to assemble the tip to the spring. This extra bonding step adds significant complexity, yield loss, and cost to the manufacturing process. Furthermore, the pyramid shape produces a contact geometry that is limited to a square or rectangular contact surface which grows in size as it is abraded or re-surfaced as is often done in practical application as a result of abrasive cleaning. Any change in surface shape or size results in a change in contact area and hence electrical contact characteristics as well as scrub mark.
Another way of solving machine vision problems is to create the tip significantly tall (approximately 50 um tall). In this embodiment, the next underlying planar surface (the post) would be far enough away from the focal plane of the vision microscope so that the post surface would be out of focus and only the tip would be in focus. However, this is not a practical solution for tips that are produced by lithographic imaging and electroforming. Such processes have practical limits in aspect ratio (height to width ratio). Furthermore, even if the aspect ratios of a taller tip were practical (typical tips are about the same height as or slightly higher than their smallest dimension which is on the order of 5 um to 20 um), a taller tip would be prone to breakage from the lateral scrubbing forces present in use.
Another proposed alternative is to remove part of the post structure, creating a sloped surface around the tip, see FIG. 5B, that cannot reflect illumination back to the vision system. However, a problem with this design is that it is very difficult to align a tip on the flat top of the now tapered post. Any slight misalignment provides a planar reflective surface near the tip base and causes a bright “crescent” to appear around the tip. The crescent effect interferes with proper tip-position recognition causing vision “rejects” or errors in the captured visual centroid.
Thus a new design is needed for creating a tip and post structure that will resolve the issues of vision errors when a tip is lithographically formed on a probe structure.